Biotecnologia Aplicada Elfos Scientiae ISSN: 0684-4551 Vol. 17, Num. 1, 2000, pp. 46-47

Ruedi Aebersold, Beate Rist, Steven P Gygi

Department of Molecular Biotechnology, University of Washington, Seattle, WA 98195, USA.

With the completion of a rapidly increasing number of complete genomic sequences much attention is currently focused on the questions if and how the information contained in sequence databases can be interpreted in terms of the structure, function and control of biological systems. Quantitative proteome analysis, the global analysis of protein expression, has been proposed as a method to study genes at a steady state and after perturbation-induced changes. Here we discuss the justification for gene expression analysis at the protein level, highlight the limitations in the current standard proteome technology, and introduce a new experimental approach to quantitative proteome analysis.

The poor correlation between mRNA and protein levels in cells provides justification for quantitative proteome analysis. With recent technical advances including the development of differential display-PCR [1], cDNA microarray and DNA chip technology [2, 3], and serial gene analysis (SAGE) [4, 5], it is now feasible to establish global and quantitative mRNA expression maps of cells and tissues in species for which the sequence of all the genes is known. The discoveries of post-transcriptional mechanisms which control translation rate [6] and protein and mRNA half-lives [7] led us to predict that quantitative transcript expression measurements are insufficient for predicting the quantity of protein expression. To test this hypothesis we determined the correlation between the mRNA and protein levels for a group of genes expressed in exponentially growing cells of the yeast Saccharomyces cerevisiae. Protein expression levels were quantitated by metabolic labeling of the yeast proteins to a steady state, followed by 2D-gel electrophoresis and liquid scintillation counting of the selected, separated protein species. Separated proteins were identified by tryptic digestion of spots with subsequent analysis by microcapillary high performance liquid chromatography-tandem mass spectrometry (mLC-MS/MS) and sequence database searching [8-10]. The corresponding mRNA transcript levels were calculated from serial analysis of gene expression (SAGE) frequency tables [5].

The correlation between mRNA and protein levels was calculated for a data set consisting of more than 100 mRNA and protein products of selected genes. For the entire set of genes, there was a general trend of increased mRNA levels resulting in increased protein levels. The Pearson product moment correlation coefficient for the whole data set was 0.935. This number is highly biased by a relatively small number of genes with very large protein and message levels. A more representative subset of the was the group of genes for which message level was measured below 10 copies/cell. This subset included 70% of the data used in the study. The Pearson product moment correlation coefficient for this data set was 0.356. This weak correlation is further evident by the observation that levels of protein expression coded for by mRNA with comparable abundance varied by as much as 30 fold and that the mRNA levels coding for protein with comparable expression levels varied by as much as 20 fold. This study, for the first time, correlated the mRNA transcript and protein expression levels of a relatively large number of genes expressed in cells representing the same state. lt is apparent that the observed correlation is not sufficiently high to allow for protein levels to be predicted by mRNA levels. We therefore conclude that quantitative proteome analysis is an essential component of any comprehensive analysis of biological systems.

Current proteome technology is biased towards the analysis of high abundance proteins: the current standard approach to quantitative proteome analysis is based on the separation of proteins by 2D-gel electrophoresis (2DE) and the subsequent identification of individually separated and detected protein spots by mass spectrometry or tandem mass spectrometry followed by sequence database searching [9-11]. The method is sequential, labor intensive and difficult to automate. lt does, however, provide precise quantitation and is well suited to reveal relative changes in protein expression, clusters of concurrently regulated proteins and additional features which affect the electrophoretic mobility of proteins, including post-translational protein processing and modifications. As a true proteome technology, the 2DE/MS/MS method would be expected to display every protein in a protein mixture. To assess to what extent the 2DE protein pattern obtained from a total yeast lysate represented the proteome of this microorganism, we related protein expression levels from protein detected by silver staining to the predicted expression levels of all the open reading frames (ORF) in the yeast.

Prediction of the level of protein expression was based on the codon bias of the respective genes. The codon bias indicates the propensity for a gene to utilize the same codon to encode an amino acid even though other codons would insert the identical amino acid into the growing polyeptide chain. Its value varies between 960.3 and 1.0, and it has further been found empirically that highly expressed proteins have large codon bias values (>0.2) and proteins expressed at low levels have low codon bias values (>0.1) [12]. Comparison of the codon bias distributions for all the yeast ORFs with the distribution of all the proteins analyzed by 2DE, silver staining and tandem mass spectrometry indicated that the population of proteins analyzed by the standard 2DE/MS/MS proteome analysis technique was highly biased towards the most highly expressed proteins. No proteins with codon bias values <0.2 were detected, whereas the majority of ORF92s predicted from the yeast genome sequence have codon bias values <0.2. We therefore conclude that the current proteome technology, used without sample pre-enrichment is not a true proteome technology and that the construction of complete proteome maps will be very challenging, even for relatively simple, unicellular organisms.

A novel method for quantitative proteome analysis

To address the limitations inherent to the 2DE/MS/MS method to proteome analysis, we have developed a new experimental approach. lt is intended to retain relative quantitative information while still rapidly and conclusively identifying even the minor components of a mixture. This method is based on a class of new chemical reagents termed isotope coded affinity tags (ICAT) and MS/MS. ICAT reagents consist of three functional units, namely a chemical reactivity directed towards a functional group in proteins (e.g. SH, NH2, COOH), a linker group synthesized in isotopically heavy and light forms, respectively, and an affinity tag (typically a biotin group) [13].

The ICAT strategy consists of the following steps. Proteins in protein mixtures 1 and 2 are treated after reduction with a sulfhydryl-specific ICAT reagent. The reagents exist in two forms: isotopically light (d0) and isotopically heavy (d8). The heavy and light forms are used to derivatize the proteins in samples 1 and 2, respectively. After treatment with the ICAT reagents the samples are mixed, At this point, any optional fractionation technique can be performed to enrich for low abundance proteins or to reduce the complexity of the mixture, while the relative quantities are maintained. The combined protein sample is then proteolyzed and the ICAT-tagged peptides are selectively enriched by avidin-biotin affinity chromatography. These peptides are separated and analyzed by microcapillary HPLC-ESI-MS/MS. The relative ion intensities of the two differentially isotopically tagged forms of a specific peptide indicate their relative abundance. Such pairs of tagged peptides are easily detected because they essentially co-elute from the column and because of the eight mass units difference encoded in the ICAT tag, which is detected in the mass spectrometer. Every other scan is devoted to fragmenting and then recording sequence information about an eluting peptide (MS/MS spectrum). The protein from which this peptide originated is then identified by searching a sequence database with the recorded MS/MS spectrum. The procedure thus provides the relative quantitation and identification of the components of protein mixtures in a single analysis.

In this manuscript, we argue that in the emerging post-genomic era technologies that can quantitatively, globally, and automatically measure gene expression at the protein level are essential for the comprehensive analysis of biological processes and systems. We indicate the limitations of the current standard method for large scale protein analysis with respect to the analysis of low abundance proteins and propose a new approach to quantitative proteome analysis. We anticipate that the new ICAT strategy will provide broadly applicable means for the quantitative cataloging and comparison of expressed proteins in a variety of normal, developmental, and disease states.

Acknowledgments

This work was supported in part by the NSF Science and Technology Center for Molecular Biotechnology, NIH grant T32HG00035 and a grant from the Merck Genome Research Institute.

References

1. Liang P, Pardee AB. Science 1992; 257:967.

2. Shalon D, Smith SJ, Brown PO. Genome Research 1996;6:639.

3. Lashkari DA, DeRisi JL, McCusker JH, Namath AF, Gentile C, Hwang SY, et al. PNAS USA 1997;94:13057.

4. Velculescu VE, Zhang L, Vogelstein B, Kinzler KW. Science 1995;270:484.

5. Velculescu VE, Zhang L, Zhou W, Vogelstein J, Basral MA, Bassett DE, et al. Cell 1997;88:243.

6. Harford JB, Morris DR. Post-transcriptional gene regulation. Wiley-Liss, lnc., New York, 1997.

7. Varshavsky A. PNAS USA 1996;93: 12142.

8. Eng J, McCormack AL, Yates JR. J Am Soc Mass Spectrom 1994;5:976.

9. Gygi SP, Rochon Y, Franza BR, Aebersold R. Molecular and Cellular Biology 1999; 19:1720.

10. Gygi SP, Han DKM, Gingras AC, Sonenberg N, Aebersold R. Electrophoresis 1999;20:310.

11. Patterson SD, Aebersold R. Electrophoresis 1995;16:1791–814.

12. Bennetzen JL, Hall BD. J Biological Chem 1982;257:3026.

13. Gygi SP, Rist B, Gerber SA, Turecek F, Gelb MH, Aebersold R. Nature Biotechnology In press. (1999).

Copyright 2000 Elfos Scientiae